Académique Documents
Professionnel Documents
Culture Documents
DOI: 10.1038/ncomms1891
Nanotechnology, with its broad impact on societally relevant applications, relies heavily on the
availability of accessible nanofabrication methods. Even though a host of such techniques exists,
the flexible, inexpensive, on-demand and scalable fabrication of functional nanostructures
remains largely elusive. Here we present a method involving nanoscale electrohydrodynamic
ink-jet printing that may significantly contribute in this direction. A combination of nanoscopic
placement precision, soft-landing fluid dynamics, rapid solvent vapourization, and subsequent
self-assembly of the ink colloidal content leads to the formation of scaffolds with base diameters
equal to that of a single ejected nanodroplet. The virtually material-independent growth of
nanostructures into the third dimension is then governed by an autofocussing phenomenon
caused by local electrostatic field enhancement, resulting in large aspect ratio. We demonstrate
the capabilities of our electrohydrodynamic printing technique with several examples, including
the fabrication of plasmonic nanoantennas with features sizes down to 50nm.
1 Laboratory of Thermodynamics in Emerging Technologies, Institute of Energy Technology, Department of Mechanical and Process Engineering,
ETH Zrich, CH-8092 Zrich, Switzerland. 2 Laboratory of Physical Chemistry, ETH Zrich, CH-8092 Zrich, Switzerland. Present address:
Max Planck Institute for the Science of Light, 91058 Erlangen, Germany. Correspondence and requests for materials should be addressed to
D.P. (email: dimos.poulikakos@ethz.ch).
nature communications | 3:890 | DOI: 10.1038/ncomms1891 | www.nature.com/naturecommunications
ARTICLE
Results
Working principle and set-up. Figure 1a displays the essential
steps in the temporal sequence of ENA NanoDrip printing. In the
first step, a voltage is applied between the pipette/ink and a counter
electrode, resulting in the generation of a meniscus and periodic
ejection of ink nanodroplets from its lowest point. Once ejected,
the nanodroplet is accelerated by the electric field towards the
substrate where viscous effects pronounced by the nanodroplet
size facilitate a soft landing and, owing to wetting-enhanced rapid
vapourization, the droplet can dry during an ejection period e.
Once the solvent has vapourized, the nanoparticles, formerly
dispensed in the ink droplet, are found to be uniformly spread in an
area corresponding to the projection of the spherical nanodroplet
(Fig. 1b). Periodic occurrence of these events leads to a narrow
accumulation of nanoparticles on the substrate. As this structure
grows in height (Fig. 1c,d), its tip acts as a sharp electrode, creating
strong electric-field gradients that focus the incoming droplets
towards it and allow the structure to grow at a homogeneous
diameter. The interval between voltage initiation and termination is
denoted as pulse, as it has a defined duration and a sharp rise and fall
(Fig. 1a). However, the pulse length does not influence the ejection
frequency, nor the ejection process in general. The only purpose
of using pulses is to precisely control the number of deposited
droplets, that is, nanoparticles, and therewith the extent of printed
ARTICLE
Applied voltage
Time
Printing stage
0V
XYZ micro-stage
Micropipette
Substrate
Counter electrode
XYZ piezo-stage
Microscope
Control unit
Electroactuation
HV amplifier
iSCAT
400 Vp
Figure 1 | Schematic of ENA printing process and set-up. (a) Growth of a liquid meniscus and subsequent ejection of ink nanodroplets from its apex on
application of a DC voltage. During DC on-time, droplets are ejected at a homogeneous period e and, once impacted, are vapourized (represented by
wavy arrows) in the course of e. After periodic repetition of this event (for illustration convenience merged into one cycle), a sharp structure consisting of
a multitude of formerly dispersed nanoparticles rises from the substrate, attracting approaching charged droplets by ENA (straight arrows). The growth
process is further illustrated with SEM micrographs (150nm scale bar) of (b) the deposition pattern of a single nanodroplet and that of (c,d) actual
nanopillars. (e) Schematic of the ENA NanoDrip set-up with the nozzle located above an underlying glass substrate placed on an ITO-coated glass slide
representing the grounded counter electrode. Voltage stimuli were applied in the form of amplified DC signals between the ink-filled, metal-coated pipette
and the counter electrode. A homebuilt iSCAT imaging system was used to detect printed structures.
ARTICLE
10
1
100
0.1
100
150
200
250
300
Applied voltage (V)
b
100
10
1
100
150
200
250
300
350
0 V m1
5e7
1e8
1.5e8
2e8
2.5e8
Z
3e8
300
200
Applied voltage (V)
100
Nondimensional droplet
diameter
100
350
1,000
100
10
1
0.1
0.1
1.0
1.5
2.0
2.5
Nondimensional voltage
100
Diameter (nm)
1,000
Diameter (nm)
3.0
ARTICLE
0 Q0
1 Q0
2 Q0
Figure 4 | Electrostatic field-focussing effect for charged and uncharged nanopillars. The images show nanopillars subjected to an applied electric
field. The arrows point to the direction of acceleration of a charged nanodroplet with an arbitrary intensity represented by the arrow length. The figure
compares the resulting field for three structures, bearing a residual charge of (a) 0, (b) 1 or (c) 2 droplets (Q0=unit droplet charge). Because residual
charge acts repulsively to new approaching droplets, the field-focussing effect is overtaken by repulsion already for one residual droplet charge.
This illustrates that also from the charging point of view, complete vapourization between consecutive droplet ejections is a requirement for ENA
NanoDrip printing. Scale bar, 50nm.
volume. At least for a droplet of the size of the meniscus, this may
be understood by scaling the flow rate with the help of the classical
Poiseuille equation (Supplementary Methods).
Nanostructure growth initiation. The cyclic release and deposition
of spherical droplets, one at a time, in nanodripping allows enough
time for the efficient removal of solvent by vapourization during e.
This lets nanoparticles of a multitude of footprints accumulate into
a tight scaffold, which only increases in height while maintaining
constant lateral size. This mechanism is in great contrast to that of
cone-jet printing in which the ink and the accompanying nanoparticles continuously spread on the substrate owing to the high immediate volume flow rates7,8.
Fast vapourization of a droplet implies that to avoid clogging at the nozzle, the volumetric rate of vapourization at a single
deposited droplet must be higher than that at the meniscus itself
(Supplementary Methods). Our experiments show that this is fulfilled, if the ink is able to sufficiently wet the substrate surface14.
Indeed, the employed glass slips exhibited excellent wetting properties with contact angles <10. In addition, we registered very strong
wetting also when depositing optically observable volumes of solvent onto dried nanoparticle layers. This behaviour is expected due
to the solvent-specific coating of the nanoparticles. In contrast to the
outcome on wetting substrates (that is, nanopillars with homogeneous diameter), the deposition of droplets onto only partially wetted
substrates (contact angle ~60; see Methods) resulted in nano
particle scaffolds with spherical cap geometry several times larger
than a single droplet (Supplementary Fig. S7). Such a shape is clear
evidence of a single drying sessile ink droplet containing mobile
nanoparticles15 and illustrates the accumulation of several smaller
droplets on the substrate due to insufficient vapourization rates.
An interesting and important feature of ENA NanoDrip printing
is the lack of a coffee-stain effect16 and the equality of droplet and
footprint size despite strong wetting of the ink. The lack of a coffeestain pattern can be explained by the low Peclet number in the range
of unity17, giving rise to diffusion-dominated nanoparticle motion
(Supplementary Methods). However, the small size of footprints
requires further studies that go beyond our current work. We believe
that the underlying mechanism of this unexpected behaviour is that
size-related strong viscous damping causes a soft-landing process despite the high impact velocities (Supplementary Fig. S8 and
Supplementary Methods) whereas electric forces are not found to
notably enhance impact spreading (Supplementary Table S1 and
Supplementary Methods). Hence, a droplet will initially be deposited
on the substrate with a diameter similar to that of its spherical state,
and only afterwards, capillarity-driven wetting will slowly increase
its diameter18 (Supplementary Methods). The observed equality of
droplet and footprint sizes implies that, in our case, nanoparticle
settling occurs before the longer wetting phase. However, since the
time that a nanoparticle needs to diffuse along the droplet domain is
similar to the expected wetting time, diffusion alone is not expected
to allow full nanoparticle settling during the required timescale
(Supplementary Table S2 and Supplementary Methods). We speculate that because of the small size of the deposited droplets, longrange particle-substrate forces, especially Coulomb forces, extend
over the entire droplet domain, thus confining the particle settlement process17,19. The simultaneous presence of ionic charge and
external electric field renders rigorous quantification difficult and
requires further study. We note in passing that footprint patterns
did not differ for nanodroplets of opposite charge.
Structure growth by electrostatic nanodroplet autofocussing.
Once the first few droplets have laid the foundation of a pillar
(Fig. 1a), the electrostatic field induced at its strongly curved surface focusses the following incoming nanodroplets to the tip of
the forming pillar. Because of the zero lateral electric field component at the nozzle-substrate axis, droplets stay in a narrow path
even when working at nozzle-substrate distances as high as ~5m
(further increase in this separation can eventually lead to unfocussed
structure growth; see Supplementary Fig. S9). Figure 4ac show the
results of electrostatic simulations for three scenarios of different
residual ionic charge originating from the deposited charged droplets. The figures suggest that electrostatic autofocussing only acts on
droplets in close proximity to a growing nanostructure and only if
the deposited charge is sufficiently removed. Figure 4a represents
the most optimal situation of full charge removal during the course
of solvent vapourization. In this case, the metal structure clearly
acts attractive on approaching nanodroplets. Simulation results for
residual charge, equal to that carried by 1 droplet and 2 droplets,
respectively, (Fig. 4b,c, respectively) imply that, on a non-conductive
substrate liquid- and concomitant-charge accumulation would
lead to detrimental repulsion effects between pillar structure and
approaching charged droplet. Thus, observation of high aspect ratio
pillars, as that shown in Fig. 5a with 50-nm width and 850-nm
height, confirms the need for rapid liquid vapourization carrying
away ionic charge. In combination with the processes enabling uniform footprint patterns at the same size as a droplet, ENA leads to
an equality of droplet and pillar diameter. We verified this important feature by analysing the diameters of nanopillars printed at a
set of voltages and plotting them together with measured footprint
sizes (Supplementary Fig. S10).
The combination of small droplet trajectory divergence and electrostatic autofocussing allows not only a remarkable aspect ratio in
the formation of nanostructures, but also a superb spatial control of
structural placement as illustrated by an array of printed ~80-nm
wide nanodots of 1m lattice constant (Fig. 5b). The deviation
ARTICLE
Height (nm)
40
20
0
0
Figure 5 | SEM micrographs of printed nanostructures with dimensionalities from 0D to 2D. (a) Gold nanopillar of diameter ~50nm and aspect ratio
of ~17 (Scale bar, 200mm). (b) Top and (c) side view of nanopillars printed subsequently at 200nm center-to-center distance (scale bar, 200nm).
(d) 80-nm wide dots printed into a 1-m lattice constant array (1m scale bar). (e) Printed tracks with pitch sizes of 250, 200, 150, 100 and 75nm
(scale bar, 2 m). The inset shows AFM (full black lines) and SEM (red dashed lines) profiles of 150-nm pitch size. The height of AFM profiles is
given in nanometers. The SEM profiles are in arbitrary units. Tracks have reproducible heights of ~40nm and are well separated.
between the actual and aspired positions is in the range of only a few
nanometers. The spacing between structures can be further reduced
as shown by a pair of sequentially printed nanopillars of diameter 120nm, separated by a gap of 80nm (Fig. 5c,d). This example
emphasizes that moving the central impact region by an amount as
small as the structure diameter is sufficient to prevent droplets from
being attracted to the previously printed nanopillar.
The attainable positioning accuracy and resolution are further demonstrated in Fig. 5e with an example of printed in-plane
tracks. These were created by introducing a constant relative movement between the nozzle and the substrate during droplet ejection.
Figure 5e displays tracks at pitch sizes ranging from 75nm to
250nm, where full separation is achieved for track distances down
to 100nm. Atomic force microscopy profiles further reveal reproducible track heights of about 40nm (Fig. 5e, inset). Such tracks may
be employed as electrical conductors. Supplementary Figure S11
provides an example of a printed track after an annealing treatment,
proving that those structures do not loose integrity when being
sintered but instead form one entity. Quantitative results on
the electrical behaviour of printed tracks in relation to different
annealing treatments will be dedicated to future studies.
Flexibility towards growth direction and material. Interestingly,
ENA also allows fabrication of nanopillars at a variety of tilting
ARTICLE
1.71 m s1
34
2.95 m s1
expected 0
1.0
0.8
Intensity (a.u.)
1
0.75 m s1 1.23 m s
50
65
0.6
0.4
0.2
Discussion
In this work, we have presented feature sizes down to 50nm by using
nozzle diameters in the range of 1m. Downscaling these by only
a factor of 25 would make our method fully competitive with the
state-of-the-art performances of EBL or FIB. Observed footprints,
originating from a 600-nm nozzle, have already reached ~35nm.
Besides reducing the size of employed nozzles, even smaller droplets could be achieved by pinning the meniscus not at the outer,
but at the inner opening30, for example, by treating the nozzle
0.0
500
550
600
650
700
Wavelength (nm)
750
800
surface with a solvent-repelling coating1,30. This may allow droplets as small as 20nm from a 600nm nozzle at a manageable effort.
Further downscaling may be achieved by employing ever smaller
nozzles. Along this line, we believe that a lower limit to the process
will mainly be induced by difficulties of further nozzle downscaling
(also the involved clogging issues) or size restrictions imposed by
the nanoparticles (as particles should still be substantially smaller
than the structures they build). Of great importance for any technical advancement will also be a better understanding of the different
physical aspects. We have identified several stringent conditions,
which have to be fulfilled for proper functioning of ENA NanoDrip
printing. Most important is a proper adjustment of the diverse flow
rates, namely of ejection flow rate and vapourization flow rate at
both, droplet and meniscus (see also the Supplementary Methods,
concerning this topic). The implementation of strongly deviating
solvents (for example, with respect to surface tension, viscosity,
vapour pressure and so on) will not be successful at the same conditions used here. In this respect, it will be of importance to also
include external factors that we have not yet investigated (for example, temperature). We believe that both downscaling and further
process versatility will be accompanied with a better understanding of the process. In contrast to techniques like EBL and FIB, ENA
NanoDrip printing is orders of magnitude less expensive, does not
require vacuum, and can easily create out-of-plane structures. We
anticipate that availability of reproducible microfabricated print
ARTICLE
heads31 in the future will introduce further controllability and parallelization of the process with a multitude of nozzles, providing an
on-demand, bottom-up and scalable nanofabrication technique.
diameter D). The charge relaxation time describes the process of relaxing charges
on a conductor surface, that is, the time it takes for the surface potential to settle
after a voltage stimulus and is defined as
te =
Methods
Printing process. The set-up shown in Fig. 1a consists of an in-house built micro
scope equipped with a 3D piezo-stage (MadCityLabs), electrical equipment for
pulse generation and a central control unit. A glass substrate was placed on top
of an ITO coated glass slide mounted at the piezo-stage. The piezo-stage had a
working distance of 300m. In the respective area, structures can be well aligned
at the inherent precision of the piezo-stage (generally <10nm). For optical detection, this microscope worked in the iSCAT mode32. This allowed highly sensitive
detection of printed nanostructures and tight focussing (low view of depth). The
latter was essential for controlling the nozzle-substrate distance (see below for
more information). For the respective experiments, the wettability of the substrate
with respect to the Au ink was reduced by vapour phase coating of a self-assembled
monolayer of 1H,1H,2H,2H-perfluorooctyltrichlorosilane, resulting in a contact
angle of ~60.
Glass pipettes with outer diameters between 550nm and 1,300nm were
prepared with a Sutter P-97 pipette puller. Only pipettes with good nozzle quality,
that is, having no cracks, chipped edges or the like, were used. Ink was introduced
to the backside opening of the pipette and drawn to the very tip by capillary forces
without any applied pressure. The outer pipette wall and a small region of the
nozzle tip interior were coated with a 10-nm adhesion layer of titanium followed
by 100nm of gold using e-beam evaporation. Irrespective of the small nozzle size,
only minimal caution was necessary when handling them. The nozzle-substrate
distance was assessed with a precision of about 150nm by tightly focussing at the
capillary tip and then moving the focal plane at a controlled distance by a movement of the micro-stage holding the objective. The substrate was then placed into
focus by a z-movement of the piezo-stage. For most experiments, the employed
ink contained 37nm gold particles in cyclododecene (ULVAC Technologies). The
ink was diluted by a factor of 50 with n-tetradecane to arrive at a final concentration of ~0.1 vol%. Dilution was necessary to avoid clogging, at least for a period of
several minutes. At substantially higher concentrations, rapid clogging prohibited
a smooth workflow. Early clogging could normally be reversed by the application
of a high-voltage pulse (up to 400V). For calculations, the physical properties
of n-tetradecane were adopted. The measured conductivity of the diluted ink
was 6.8108Sm1. Silver (ULVAC Technologies) and zinc oxide (Nanograde)
inks were both based on n-tetradecane and contained nanoparticles with sizes of
37nm and ~25nm, respectively. Voltage signals were generated by a waveform
generator (RIGOL) and amplified by a homemade HV amplifier to up to 400Volts.
Annealing and optical measurements. Thermal annealing was performed in a
rapid thermal annealing oven (Jipelec). During purge-pumping by nitrogen gas,
the temperature was first ramped at 2Cs1 to 100C, then at 10Cs1 to 200C
and finally at 3Cs1 to 260C. Once the temperature reached 260C, the heat
source was deactivated and a 5% mixture of H2/N2 was introduced at 500s.c.c.m.
The sample was passively cooled to room temperature. Scattering spectra were
obtained in a low grazing angle arrangement (Supplementary Fig. S14).
Requirements for the nanodripping mode. To achieve the nanodripping mode,
the characteristic time of supply of liquid to the droplet (q) has to be longer than
the characteristic time of drop formation11. Here q is estimated as the duration
between consecutive droplet ejections (e) and therefore equals the droplet volume
(~d3) divided by the volume flow rate V .
tq = te =
pd3
6V
(1)
td =
md
g
(2)
The drop formation time states how long it takes for a liquid element, ejected
through a circular exit with diameter d, to form a droplet solely by the action of
surface tension and viscous resistance. If the characteristic time of supply of liquid
is shorter than the drop formation time, the ensuing convective liquid flow will not
allow for the relaxation of the liquid element into a droplet. Instead, a continuous
jet can form, which would be the basis of the cone-jet ejection mode.
A timeframe right before droplet ejection is schematically depicted in Fig. 2c,
where the meniscus diameter is assumed to be equal to the outer nozzle diameter.
During ejection, the meniscus has to be quasi-steady, which requires the charge
relaxation time to be smaller than the time of supply of liquid to the meniscus
Q (the latter is also given by equation (1) but with d replaced by the meniscus
e
s
(3)
Here and are the liquid electric permittivity and conductivity. If the charge
relaxation time is longer than the time of supply of liquid, more charge would
be removed from the surface by liquid flow than what is delivered by bulk
electrical conduction. The surface charge would therefore diminish causing the
meniscus to inherently collapse. In Supplementary Fig. S15 a distribution of
these important timescales is shown, proving that the requirements leading
to nanodripping ejection are fulfilled for the entire regime investigated in
this study.
Numerical simulations. Numerical simulations for the determination of the
electric field were performed by COMSOL multiphysics software in the AC/DC
electrostatics module, using the linear UMFPACK solver. The geometry of the
pipette was determined from SEM micrographs and modelled as axisymmetric.
The model of the nozzle region includes the pipette, a pulled liquid meniscus and
a small pendant droplet at its tip, all of which are assumed to be at equipotential.
The grid was automatically refined until results deviated by less than 2%. To
estimate the droplet diameter, d, the following balance of surface tension and
electric forces has to be fulfilled:
e0
E 2 (d ,V ) daz = p dg
2
(4)
Here represents the surface tension, E is the electric field, the permittivity of air
and daz a differential surface element for the integration over the small hemispherical surface with subscript z denoting the z-direction. A surface plot of the obtained
electric field can be found in Fig. 2c. The integral in equation (4) was evaluated by
solving for the Maxwell stress tensor in the z-direction followed by integration on
the surface of the small pendant droplet. Input voltages were iteratively adjusted
until the resulting electric force matched the surface tension force with an accuracy
of better than 2%.
References
ARTICLE
32. Lindfors, K., Kalkbrenner, T., Stoller, P. & Sandoghdar, V. Detection and
spectroscopy of gold nanoparticles using supercontinuum white light confocal
microscopy. Phys. Rev. Lett. 93, 037401 (2004).
Acknowledgements
Financial support for the work reported in this paper provided by the Swiss National
Science Foundation through grant number 2-77485-09 is gratefully acknowledged. The
authors would also like to thank Dr. M. Tiwari of the Laboratory of Thermodynamics in
Emerging Technologies of ETH Zurich and Dr. Mario Agio of the Nano-Optics group of
ETH Zurich for useful discussions.
Author contributions
P.G. and J.S. built the set-up, designed and analysed experiments. P.G. interpreted data
and performed FEM simulations. P.G. and S.K. designed the model for the simulation
and performed experiments. P.G., D.P. and V.S. wrote the manuscript, which was edited
by all authors. H.E. designed and performed the optical measurements and interpreted
optical spectra. D.P. supervised all aspects of the project, designed the research and gave
scientific and conceptual advice. V.S. supervised the optical measurements and gave
scientific advice.
Additional information